Performance regression on travis

[Byron]

Py3 runs are 3 to 6 times slower on travis, even though this doesn't show on my machine at all.
What's up with that ?

[Byron]

It seems that all that time is spend in rev-parse. The latter might trigger plenty of file accesses, but it's still the question why this is so much slower in py3 all of the sudden.

[Byron]

For me, this issue isn't reproducible, neither on OSX (native), nor on an ubuntu64 VM. This is the VM's result:

[Byron]

Performance regression is still visible on travis, even though said rev-parse test was nerfed to the point where we save 2 or 3 minutes at least. However, there must be more slow tests that need to be handled somehow.

[Byron]

This seems to be gone !